Network Function Virtualization – Putting the Pain on Traditional Databases
A growing number of NuoDB customers are key players in the mobile telecommunications eco-system market. And the big story for them is Network Function Virtualization (NFV). No more hardwired components; routers, switches, and firewalls are being turned into software that runs somewhere in a cloud. Well, not just a cloud but specifically a telecom-grade cloud.
In 2015, this segment has grown into a $2B+ market and is set to grow five-fold in the next four years. The benefits of this shift are focus on a dramatic cost reduction, simplification, and innovation for the cutthroat telecom business. There is huge investment by suppliers and massive commitment from adopters. Telefonica CTO Enriques has said recently, “We can’t survive if we don’t do this”. AT&T’s SVP of Architecture, Andre Fuetsch, has a very ambitious goal “To virtualize and control over 75% of our (AT&T’s) target network under our Domain 2 architecture by 2020”.
But these guys, despite their commitment, know it’s going to be a long haul over a rocky road.
While outcomes are still the same for the consumer (a phone call, a usage bill, an email received), the ways these are accomplished by the provider are very different. Applications and services are likely to remain the same, but providers will need to scale massively as their consumers’ appetite for more data and more services continues to grow.
And the environment in which they exist will be very different. Instead of hardwired devices we’ll have software running on virtual hardware. Instead of appliances and dedicated resources in data centers supporting services, we’ll have virtual resources distributed across the networks. So the challenge is to build flexible infrastructure in this virtual world that can run existing services and support virtualized components – but with limitless scaling and the support for cool new services and functions.
Much of this is already being proven in more general cloud deployments that are delivering:
- Commodity resources – basic computing resources (compute, memory, network, and storage) are delivered by virtualized commodity hardware
- Scale-out elasticity – scale is achieved by distributing workloads across many such commodity resources
- Geographic and global distribution – processing occurs in multiple locations at once and co-operating resources may be in data centers anywhere around the globe
- Low-touch operations – the lifecycle of every resource is automated. Managed resources needs minimal oversight and no customization.
Cloud architectures have already shown themselves to be very successful for many use cases. Processing individual user requests is easily distributed and scaled, as is serving the application context for vast numbers of interactions.
But at the back of such systems usually lies a database, typically a relational database (RDBMS). And a traditional RDBMS only scales upwards, by running on a bigger and bigger, faster and faster machine. That becomes very expensive as you move into the realms of specialist, high-performance hardware.
This approach is also fundamentally at odds with the emerging cloud and virtualized architectures that exploit a scale-out model of sharing workload across as many commodity boxes as necessary. And RDBMSs struggle with distributed processing. Database sharding, replication, real-time change data capture (CDC), and a host of other complex point solutions have been deployed which even then, and even in the hands of dedicated experts, still don’t fully meet the demands of global operations.
That is the problem NuoDB set out to solve, how to keep the value of RDBMS - rich SQL programming model and transactional consistency - but with scale-out and geo-distribution.
We have cracked this very important problem with our unique new database architecture, which is probably why our story has so much resonance for many of the players in the world of NFV.
Boris Bulanov is Vice President of Technology at NuoDB.